How to draw gene expression heatmap ?

Question

[takoyaki]

I want to draw heatmap using differential expression genes.

First, I mapped using hisat2, and then counted genes using htseq. From this expression count matrix, I used DESeq2 to detect DEGs.

Now I have DEG list. I want to draw heatmap about these DEGs.

Which expression value do many people use to draw heatmap ? In my pipeline (hisat2 → htseq-count → DESeq2), which data can be used to draw genes expression heatmap ?

[Bastien Hervé]

Depends of what you want to show. Usually heatmap are made on the log of the counts

http://bioconductor.org/packages/devel/bioc/vignettes/DESeq2/inst/doc/DESeq2.html#heatmap-of-the-count-matrix

Something like :

pheatmap(log2(counts(dds, normalized=TRUE))...)

[takoyaki]

Thank you for your reply !　I appreciate you sharing URL. I will use it as a reference.

Answer 1

[Kevin Blighe]

You just need to subset your normalised expression data (in DESeq2, it is returned by counts(dds, normalized=TRUE), as pointed out by Bastien), but you can also use the normalised and then transformed expression data - these are produced via rlog() or vst(), then accessed via assay(). For example:

rlogcounts <- rlog(dds, blind = FALSE)
dataHeatmap <- assay(rlogcounts)

There is no standard for what to actually use in a heatmap. You can just use whatever you want. I have seen a wide range of things:

• binary data
• raw counts
• normalised counts
• log-normalised counts
• normalised, transformed counts
• log [base 2] ratios
• Z-scores

You just need to specifically state what it is that you are using.

There is a wealth of information at your fingertips via a search engine, too.

Kevin

[takoyaki]

Thank you for your polite reply. So, there are no uniform indicator. I'm going to try some scores. Thank you !